[MDEV-8133] ALTER TABLE can perform the operation but escape the binary log Created: 2015-05-11  Updated: 2016-02-15  Resolved: 2016-02-15

Status: Closed
Project: MariaDB Server
Component/s: Data Definition - Alter Table, Replication
Affects Version/s: 10.0, 10.1
Fix Version/s: 10.0.24, 10.1.12

Type: Bug Priority: Major
Reporter: Elena Stepanova Assignee: Sergei Golubchik
Resolution: Fixed Votes: 0
Labels: None

Sprint: 10.0.22, 10.0.24

 Description   

If ALTER TABLE is interrupted, it can happen that it performs the operation, e.g. adds a column, but is not written to the binary log, which of course causes further discrepancy.

The MTR test below is for 10.1 only, because it uses max_statement_time, but the problem is not specific to it, same happens with KILL QUERY (see the RQG test which can be used for 10.0 as well).

The MTR test highly depends on the machine timing, so it might need some tuning, e.g. more or less data in the table, longer or shorter max_statement_time, etc.

MTR test case

--source include/have_innodb.inc
--source include/have_binlog_format_statement.inc
 
--disable_query_log
 
set sql_log_bin = 0;
 
create table t1 (pk int auto_increment primary key, i int, c varchar(8)) engine=InnoDB;
insert into t1 (i,c) values (1,'foo'),(2,'bar'),(3,'qux'),(4,'foobar');
insert into t1 select null, i, c from t1;
insert into t1 select null, i, c from t1;
insert into t1 select null, i, c from t1;
insert into t1 select null, i, c from t1;
insert into t1 select null, i, c from t1;
insert into t1 select null, i, c from t1;
insert into t1 select null, i, c from t1;
insert into t1 select null, i, c from t1;
insert into t1 select null, i, c from t1;
insert into t1 select null, i, c from t1;
 
set sql_log_bin = 1;
 
let $col_count = 3;
let $master_pos = query_get_value(show master status, Position, 1);
 
let $timeout = 3;
 
while ($timeout)
{
  eval set max_statement_time = $timeout;
  let $run = 10;
  while ($run)
  {
    let $colname = c$timeout-$run;
    echo # Adding $colname;
    --error 0,ER_STATEMENT_TIMEOUT,ER_QUERY_INTERRUPTED
    eval alter table t1 add `$colname` int;
    let $new_col_count = `select count(*) from information_schema.columns where table_name = 't1'`;
    echo # Error code $mysql_errno, new column count $new_col_count;
    if ($new_col_count > $col_count) {
     let $new_master_pos = query_get_value(show master status, Position, 1);
     if ($new_master_pos == $master_pos) 
     {
       show binlog events;
       show create table t1;
       echo "Column $colname was added, but ALTER statement was not written to the binlog";
       die "Binlog problem";
     }
     let $master_pos = $new_master_pos;
    }
    let $col_count = $new_col_count;
    dec $run;
  }
  dec $timeout;
}
 
drop table t1;

Example of the failure

...
# Adding c2-9
# Error code 0, new column count 15
# Adding c2-8
# Error code 1968, new column count 16
Log_name	Pos	Event_type	Server_id	End_log_pos	Info
master-bin.000001	4	Format_desc	1	248	Server ver: 10.1.5-MariaDB-wsrep-debug-log, Binlog ver: 4
...
master-bin.000001	1745	Gtid	1	1783	GTID 0-1-12
master-bin.000001	1783	Query	1	1875	use `test`; alter table t1 add `c2-9` int
Table	Create Table
t1	CREATE TABLE `t1` (
  `pk` int(11) NOT NULL AUTO_INCREMENT,
  `i` int(11) DEFAULT NULL,
  `c` varchar(8) DEFAULT NULL,
  `c3-10` int(11) DEFAULT NULL,
  `c3-9` int(11) DEFAULT NULL,
  `c3-8` int(11) DEFAULT NULL,
  `c3-7` int(11) DEFAULT NULL,
  `c3-6` int(11) DEFAULT NULL,
  `c3-5` int(11) DEFAULT NULL,
  `c3-4` int(11) DEFAULT NULL,
  `c3-3` int(11) DEFAULT NULL,
  `c3-2` int(11) DEFAULT NULL,
  `c3-1` int(11) DEFAULT NULL,
  `c2-10` int(11) DEFAULT NULL,
  `c2-9` int(11) DEFAULT NULL,
  `c2-8` int(11) DEFAULT NULL,
  PRIMARY KEY (`pk`)
) ENGINE=InnoDB AUTO_INCREMENT=8179 DEFAULT CHARSET=latin1
"Column c2-8 was added, but ALTER statement was not written to the binlog"

For running the RQG test, clone lp:~elenst/randgen/mariadb-patches (the main randgen tree might miss some changes or fixes in the required components).

RQG grammar alter.yy

query:
	ALTER TABLE _table ADD _letter[invariant] INT ; ALTER TABLE _table DROP _letter[invariant] ;

RQG command line

perl ./runall-new.pl --threads=1 --duration=200 --queries=100M --reporters=QueryTimeout,BinlogConsistency --grammar=alter.yy --engine=InnoDB --mysqld=--log-bin --rows=100,100,100,100,100,100,100,100,1000 --querytimeout=1 --basedir=<basedir> --vardir=<vardir> 

The failure can come in two flavors: either it complains about binlog replaying error due to an unknown column, or reports multiple differences like

-INSERT INTO `BB` VALUES (100,6,NULL,'2002-04-16','2002-04-16','00:43:36','00:43:36','2002-05-14 10:52:52','2002-05-14 10:52:52',NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL);
+INSERT INTO `BB` VALUES (100,6,NULL,'2002-04-16','2002-04-16','00:43:36','00:43:36','2002-05-14 10:52:52','2002-05-14 10:52:52',NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL);

where the only difference is the number of columns in the table.

If neither happens, try to increase the number of threads on the command line.



 Comments   
Comment by Sergei Golubchik [ 2015-07-27 ]

Sorry, but I cannot repeat it with rqg. Tried to increase the number of threads — didn't help. I'd rather not use the first (mtr) test case, because I need to repeat it in 10.0.

Comment by Elena Stepanova [ 2015-07-29 ]

I have set it up on perro (the RQG test).

  • login as usual

    cd mdev8133
    . ./cmd

    The server is in ~/git/10.0-mdev8133

Comment by Sergei Golubchik [ 2015-10-22 ]

I don't think this can be fixed completely. In-place ALTER cannot be undone and there are lots of actions (tables are opened, renamed, etc) that can fail after the in-place table modifications were done. After these failures we may end up with a half-done ALTER, for example, the table has the old name (if ALTER...RENAME) but a new structure.

The only “fix” I can think of is to abort the replication. This can be easily achieved by binlogging ALTER together with the error code. There will be no error on the slave and the replication will stop. Not very intuitive behavior, though.

Comment by Sergei Golubchik [ 2015-10-22 ]

Is it an upstream bug?

Comment by Elena Stepanova [ 2015-10-22 ]

I don't see any notes about it in my description or comments, apparently I didn't check. I will do it shortly.

Comment by Elena Stepanova [ 2015-10-28 ]

Could not reproduce on MariaDB 5.5, MySQL 5.5, MySQL 5.6 (for the latter, ran 20 attempts of the RQG test – for MariaDB 10.x usually 1-2 is enough).

Generated at Thu Feb 08 07:24:53 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.